How omashish/google-adk-workflows works
Project type and purpose
• This is a Python-based multi-agent trip planning system built on Google’s Agent Development Kit (ADK).
• Purpose: demonstrate how specialized AI agents (flight, hotel, sightseeing, summary) can be orchestrated by different workflow patterns (simple, dispatcher, parallel, self-critic) to solve complex travel planning tasks.
File structure
├── .gitignore
├── README.md
├── env.example ← environment template
├── requirements.txt ← dependency list
├── subagent.py ← “core agents” module (FlightAgent, HotelAgent, SightseeingAgent, TripSummaryAgent)
├── simple/agent.py ← SimpleAgent orchestration
├── dispatcher/agent.py← DispatcherAgent orchestration
├── parallel/agent.py ← ParallelAgent orchestration
├── self_critic/agent.py← SelfCriticAgent orchestration
└── (no tests directory, no microservices – all runs in a single ADK process, using modular code)
README for additional insights
• Lists four orchestration agents, four core agents, workflow patterns: sequential, parallel, feedback loops, state management.
• Mentions external dependency on Google API Key and Gemini model via ADK.
• Indicates a web interface is launched via adk web.